The ETS Family Transcription Factor ELK-1 Regulates Induction of the Cell Cycle-regulatory Gene p21Waf1/Cip1 and the BAX Gene in Sodium Arsenite-exposed Human Keratinocyte HaCaT Cells*

Cyclin-dependent kinase inhibitor (CDKN1A), often referred to as p21Waf1/Cip1 (p21), is induced by a variety of environmental stresses. Transcription factor ELK-1 is a member of the ETS oncogene superfamily. Here, we show that ELK-1 directly trans-activates the p21 gene, independently of p53 and EGR-1, in sodium arsenite (NaASO2)-exposed HaCaT cells. Promoter deletion analysis and site-directed mutagenesis identified the presence of an ELK-1-binding core motif between −190 and −170 bp of the p21 promoter that confers inducibility by NaASO2. Chromatin immunoprecipitation and electrophoretic mobility shift analyses confirmed the specific binding of ELK-1 to its putative binding sequence within the p21 promoter. In addition, NaASO2-induced p21 promoter activity was enhanced by exogenous expression of ELK-1 and reduced by expression of siRNA targeted to ELK-1 mRNA. The importance of ELK-1 in response to NaASO2 was further confirmed by the observation that stable expression of ELK-1 siRNA in HaCaT cells resulted in the attenuation of NaASO2-induced p21 expression. Although ELK-1 was activated by ERK, JNK, and p38 MAPK in response to NaASO2, ELK-1-mediated activation of the p21 promoter was largely dependent on ERK. In addition, EGR-1 induced by ELK-1 seemed to be involved in NaASO2-induced expression of BAX. This supports the view that the ERK/ELK-1 cascade is involved in p53-independent induction of p21 and BAX gene expression.

various other cellular responses, including transcriptional regulation, nuclear import, cell motility, apoptosis, DNA repair, and aging (in cellular context-and extracellular signal-dependent manners) (2,3). Aberrant up-regulation of p21 is strongly associated with cell cycle arrest, which may occur at multiple stages during the cell cycle (4 -6) and which is mediated through inhibition of the activity of cyclin E-CDK2 or -CDK4 complexes (7) or cyclin B-CDK1 (5, 8 -10) or through the degradation of cyclin B1 (11). Although p21 was initially identified as a p53 target gene, a variety of other transcription factors, including STATs, E1AF, AP-2, C/EBP, ETS-1, p150 (Sal2), Spalt, SP1, sterol regulatory element-binding protein (SREBP)-1a, and hepatocyte nuclear factor (HNF)-4␣, bind to specific cis-acting elements in the p21 promoter in response to different extracellular signals and regulate p21 expression independently of p53 (12).
Inorganic arsenic predominantly occurs in the form of either arsenite (trivalent arsenic) or arsenate (pentavalent arsenic), the two of which may be interconverted in vivo. Arsenic produces various toxic effects, including carcinogenesis, neurotoxicity, and immunotoxicity (13). A growing body of evidence suggests that chronic exposure to low levels of arsenic may be linked to the modulation of intracellular signaling pathways and gene expression profiles responsible for cell cycle progression, resulting in promotion of cell transformation (14 -16). Interestingly, cell transformation occurs only in cells exposed to low concentrations of arsenite (i.e. Ͻ5 M), whereas higher concentrations (i.e. Ն50 M) lead to apoptosis and cytotoxicity (17)(18)(19). Arsenic is a well known carcinogen in humans but has also been shown to be an effective chemotherapeutic agent (depending on cell type, arsenic species and dose, and duration of exposure) (19). Sodium arsenite (NaASO 2 ) inhibits cell cycle progression in NIH3T3 cells (20), human umbilical vein endothelial cells (21), and rat neuroepithelial cells (22) as well as in certain types of cancer cell, including SiHa cervical carcinoma (23) and A431 epidermoid carcinoma (24) cells, in all cases by up-regulating p21 expression. However, it is unclear how arsenite regulates p21 expression.
Mammals have at least five major MAPK subfamilies, of which the best known are the ERK, JNK, and p38 kinase. These major kinases play important roles in transmitting extracellular signals to cells and modulate the expression of multiple genes (25)(26)(27). Moreover, deregulation of MAPKs is associated with the pathogenesis of several human diseases (28). In general, JNK and p38 kinase are activated by growth-inhibitory signals and cellular stress, whereas ERK responds to mitogenic and cell survival signals. However, the roles of individual MAPK signaling pathways are complex. Although many studies support a role for ERK signaling in cell proliferation and survival, it has also been implicated in the transduction of antiproliferative signals in certain circumstances. For example, ERK contributes to the induction of neuronal differentiation by nerve growth factor (29) and to growth arrest and the induction of apoptosis through phosphoactivation of p53 (30,31). Elsewhere, several studies have demonstrated that ERK signaling is associated with up-regulation of p21 expression in a variety of cell types (32)(33)(34)(35)(36)(37)(38). However, despite the emerging recognition that MAPKs inhibit cell proliferation by affecting p21 expression, little is yet known about the mechanisms by which these kinases regulate p21 transcription.
ELK-1, a member of the ETS subfamily of transcription factors, is a well known substrate of ERK, JNK, and p38 kinase (39 -42). It regulates the transcription of immediate early response genes, including c-FOS and EGR-1, through serum response elements within their promoters (39 -43). Sodium arsenite (NaASO 2 ) induces the transcription of RTP801/ REDD1/Dig2, a stress response gene, in HaCaT cells by activating ELK-1 (44), suggesting that ELK-1 may contribute to the responses of HaCaT cells to NaASO 2 . Because MAPK signaling induces p21 expression in a p53-independent fashion to negatively regulate cell cycle progression and ELK-1 is a known substrate of three major MAPKs, it is possible that MAPK-mediated activation of ELK-1 may contribute to NaASO 2 -induced p21 expression. However, the functional role of ELK-1 in trans-activation of the p21 gene has not been studied. We investigated the potential role of MAPK/ELK-1 signaling in the p53-independent regulation of p21 transcription using NaASO 2 -exposed human HaCaT keratinocytes carrying mutations in both p53 alleles (45).
Here, we identified cis-acting ELK-1 response elements in the human p21 gene promoter and assessed whether ELK-1 regulates transcription of the p21 gene in HaCaT keratinocytes. We found that ELK-1 directly trans-activates the p21 gene promoter independently of p53 and EGR-1 in NaASO 2 -exposed HaCaT cells. Furthermore, we showed that the induction of EGR-1 expression by ELK-1 contributes to NaASO 2 -induced BAX expression. Based on these data, we propose an additional role of ELK-1 in mediating NaASO 2 -induced p21 and BAX expression in p53-mutated HaCaT cells.
Cell Cycle Analysis-Cellular DNA content was analyzed by flow cytometry as described previously (48). Briefly, HaCaT cells were harvested after exposure to increasing concentrations of NaASO 2 for 24 h, fixed in 70% ethanol, washed twice with PBS, and stained with a 50 g/ml propidium iodide solution containing 0.1% Triton X-100, 0.1 mM EDTA, and 50 g/ml RNase A. Fluorescence was measured and analyzed using a FACSCalibur flow cytometer (BD Biosciences).
Western Blot Analysis-Cells were lysed in a buffer containing 20 mM HEPES (pH 7.2), 1% Triton X-100, 10% glycerol, 150 mM NaCl, 10 g/ml leupeptin, and 1 mM PMSF. The resulting protein samples (20 g each) were separated by 10% SDS-PAGE and transferred to nitrocellulose filters. The blots were then incubated with the appropriate primary antibodies. Signals were developed using an enhanced chemiluminescence detection system (GE Healthcare).
Northern Blot Analysis-For each sample, 10 g of total RNA were electrophoresed on a formaldehyde/agarose gel and trans-ferred to a Hybond N ϩ nylon membrane (Amersham Biosciences). Northern blotting was performed with a [␣-32 P]dCTPlabeled p21 or EGR-1 cDNA probe, followed by hybridization with a GAPDH cDNA probe as described previously (48).
Transient Transfection and Promoter Reporter Assays-HaCaT cells were seeded onto 12-well plates and then transfected with the p21 or BAX promoter construct (0.2 g), using Lipofectamine 2000 reagent (Invitrogen), according to the manufacturer's instructions. To monitor transfection efficiency, pRL-null plasmid (50 ng), which carries a Renilla luciferase reporter, was included in all samples. Where indicated, mammalian expression vectors were also included. Then, 48 h post-transfection, cells were starved in medium containing 0.5% serum for 12 h and treated with NaASO 2 . After 6 -12 h, firefly and Renilla luciferase activities in each sample were sequentially measured using the Dual-Glo luciferase assay system. Luciferase activity in untreated cells was arbitrarily given a value of 1 (after normalization to the Renilla luciferase signal). Luminescence was measured using a Centro LB960 luminometer (Berthold Technologies, Bad Wildbad, Germany).
cis-and trans-Activation Assays-cis-and trans-activation by transcription factors was measured using the luciferase reporter assay system. To measure p53-dependent transcriptional activity, HaCaT cells were transfected with 0.2 g of cisacting reporter plasmid (13ϫp53-Luc) containing 13 tandem p53-binding sites. To measure trans-activation by ELK-1, HaCaT cells were transfected with 50 ng of trans-activator plasmid (pFA2/Gal4 DBD-Elk1), which encodes a fusion protein comprising the DNA-binding domain of yeast Gal4 (amino acid residues 1-147) and the activation domain of ELK-1 (amino acid residues 307-427), along with 0.5 g of luciferase reporter plasmid (pFR/5ϫGal4-Luc) containing five Gal4 binding elements upstream of a luciferase gene. pRL-null plasmid (50 ng) was included in all samples to allow transfection efficiency to be monitored. Following transfection, cells were treated with or without NaASO 2 and assayed for firefly and Renilla luciferase activities, using the Dual-Glo luciferase assay system.
Expression of siRNA-Short hairpin RNA (shRNA) plasmids expressing ELK-1 siRNA or scrambled control siRNA were obtained from Sigma-Aldrich. HaCaT cells were transfected with shRNA plasmids using Lipofectamine 2000 reagent (Invitrogen) or a Nucleofector device (AMAXA Inc., Gaithersburg, MD) for transient and stable expression of ELK-1 siRNA, respectively. Two days after transfection, stable transfectants were selected using G418 (400 g/ml). Knockdown of ELK-1 protein expression was verified by Western blot analysis. Generation of an expression plasmid carrying an siRNA targeted to EGR-1 mRNA (pSilencer/siEgr-1) is described elsewhere (48).
Statistical Analysis-Each experiment was repeated at least three times. Data are presented as the mean Ϯ S.D. Statistical comparisons were performed using Student's t test. A p value of Յ0.05 was considered statistically significant.

RESULTS
The Growth of HaCaT Cells Is Inhibited by Exposure to NaASO 2 -Exponentially growing HaCaT cells were treated with various concentrations of NaASO 2 for different periods of time, and the cell proliferation rate was measured. The rate of growth of HaCaT cells was markedly reduced by NaASO 2 treatment in a concentration-and time-dependent manner (Fig.  1A). A significant decrease in cell proliferation was observed in cells treated for 24 h with high concentrations of NaASO 2 (Ն50 M). Next, we analyzed cell cycle profiles in cells treated with NaASO 2 for 24 h (Fig. 1B). NaASO 2 treatment caused a slight but significant dose-dependent decrease in the G 1 population, which was accompanied by the accumulation of G 2 /M phase cells. Numbers of sub-G 1 cells, typically associated with apoptosis, also increased dose-dependently. Somewhat higher rates of cell death were observed in cells treated with 100 M NaASO 2 . Thus, treatment with NaASO 2 modulated cell cycle progression and apoptotic cell death in HaCaT cells in a dosedependent manner.
p21 and BAX Are Up-regulated in NaASO 2 -exposed HaCaT Cells-Aberrant expression of p21 is associated with growth inhibition and the induction of apoptosis in many cell types. To investigate whether NaASO 2 alters the expression of p21, HaCaT cells were treated with different concentrations of NaASO 2 , and levels of p21 protein were measured. As shown in Fig. 1C, levels of p21 increased in cells treated with NaASO 2 at concentrations of Ն50 M. In contrast, levels of cyclin D1 in cells treated with low (Յ10 M) and high (Ն50 M) concentrations of NaASO 2 increased and decreased compared with basal levels, respectively. Levels of other cell cycle-regulatory proteins, including PCNA, cyclin A1, and cyclin B1, decreased in cells treated with high doses (Ն50 M) of NaASO 2 . We further observed that NaASO 2 induced remarkable increases in the levels of BAX and cleavage of poly(ADP-ribose) polymerase (PARP), two representative markers of apoptosis, in a dose-dependent manner. When HaCaT cells were exposed to 50 M NaASO 2 for different periods of time (Fig. 1D), p21 levels increased as early as 4 h and remained elevated 24 h later. In contrast, cyclin D1 protein levels increased transiently but declined below basal levels 24 h later. Levels of cyclin A1 and cyclin B1 decreased slightly after 24 h, whereas the expression levels of BAX and cleaved PARP were detected within 6 h and thereafter gradually increased.
The cleavage of native 113-kDa PARP, yielding 89-and 24-kDa fragments could be catalyzed in a caspase-dependent or -independent pathway (53). Because NaASO 2 induces the cleavage of caspase-2 and caspase-7 in HaCaT cells (supplemental Fig. 1A), we next examined the effect of caspase  JULY 29, 2011 • VOLUME 286 • NUMBER 30 inhibitors on the cleavage of PARP. NaASO 2 -induced PARP cleavage is prevented by pretreatment with benzyloxycarbonyl-VAD-fluoromethyl ketone, a pancaspase inhibitor, or benzyloxycarbonyl-DEVD-fluoromethyl ketone, a caspase-3/-7 inhibitor (Fig. 1E). Thus, it seems likely that exposure of HaCaT cells to high doses of NaASO 2 (Ն50 M) inhibits cell cycle progression and promotes cell death via a caspase-dependent pathway. These data suggest that NaASO 2 causes cell death by up-regulating the expression of proapoptotic proteins.

ELK-1 Regulation of p21 and BAX Transcription
p53 Is Not Involved in NaASO 2 -induced Expression of p21-To determine whether NaASO 2 activates the transcription of the p21 gene, HaCaT cells were transiently transfected with a full-length p21 promoter reporter (p21-Luc(Ϫ2400/ϩ1)), and the effects of NaASO 2 on luciferase reporter activity were assessed. NaASO 2 dose-dependently increased luciferase reporter activity ( Fig. 2A). An ϳ2.9-fold increase in reporter activity was observed in cells treated with 50 M NaASO 2 (p Ͻ 0.01 versus mock-treated control). We confirmed the time-dependent induction of p21 mRNA expression by NaASO 2 using Northern blot analysis (Fig. 2B). To identify mechanisms responsible for NaASO 2 -induced p21 expression, we first investigated the role of p53. Expression of p53 protein (Fig. 2C) and p53-dependent transcriptional activity (Fig. 2D) were not significantly altered in HaCaT cells exposed to NaASO 2 . Fur-thermore, NaASO 2 induced p21 protein expression in p53-null HCT116 cells (Fig. 2E). Thus, p53 does not appear to be necessary for NaASO 2 to induce p21 expression.
EGR-1 Is Not Involved in NaASO 2 -induced p21 Expression-It has been reported that arsenite induces EGR-1 expression in HaCaT cells (54). However, the consequences of this response are unknown. We confirmed that exposing HaCaT cells to NaASO 2 activates the EGR-1 promoter (Fig. 3A). We also confirmed time-dependent induction of EGR-1 mRNA by NaASO 2 using Northern blot analysis (Fig. 3B). Treatment with NaASO 2 also causes EGR-1 protein to accumulate in a dose-dependent manner (Fig. 3C). The level of EGR-1 protein reached a peaked within 2 h after adding NaASO 2 , which gradually declined to basal levels by 12 h of stimulation (Fig. 3D). We and others have shown that EGR-1 binds directly to the proximal p21 promoter and activates p21 gene transcription (48,49,55). We used an RNA interference approach to investigate whether EGR-1 is involved in NaASO 2 -induced p21 promoter activation. HaCaT cells were transfected with an shRNA plasmid targeting specific sequence of the EGR-1 mRNA (pSilencer/siEgr-1), along with a full-length p21 promoter reporter (p21-Luc(Ϫ2400/ϩ1)). Transient transfection of EGR-1 shRNA clearly attenuated NaASO 2 -induced EGR-1 expression; however, it had no effect on NaASO 2 -induced p21 promoter activity (Fig. 3E). Furthermore, when a series of deletion mutants of the p21 promoter- driven reporter were transfected into HaCaT cells, we found that luciferase activity remained high unless the region Ϫ235 to Ϫ150 was deleted (Fig. 3F). Given that a putative EGR-1 binding sequence is present in the region Ϫ150 to ϩ38 of the p21 promoter (48 -49, 55), we suggest that EGR-1 may not be essential for NaASO 2 -induced p21 promoter activation. To fur-ther assess the involvement of EGR-1, we made point mutations in the EGR-1-binding site (gggg to tttt); these mutations had no effect on NaASO 2 -induced luciferase reporter activity (Fig. 3F). Thus, it seems likely that a NaASO 2 response element is present somewhere in the region between Ϫ235 and Ϫ150 of the p21 promoter.  JULY 29, 2011 • VOLUME 286 • NUMBER 30

ELK-1 Regulation of p21 and BAX Transcription
An ELK-1-binding Element in the p21 Promoter Is Necessary for NaASO 2 -induced p21 Promoter Activation-We next sought to identify the cis-acting element in the p21 gene responsible for NaASO 2 -induced activation. Putative transcription factor-binding sites were analyzed using the Webbased program MatInspector (Genomatix). We identified a consensus ETS-like protein-1 (ELK-1) core binding motif (TTCC; reverse complement of the commonly reported GGAA motif) between nucleotides Ϫ190 and Ϫ170 of the p21 promoter (Fig. 4A). To evaluate the role of this putative ELK-1binding element, we introduced site-directed mutations (TTCC 3 TTGG) into the core ELK-1-binding motif of the p21-Luc(Ϫ235/ϩ38) plasmid, yielding p21-Luc(Ϫ235/ ϩ38)mtElk1. The results of a promoter activity assay revealed that disruption of this core element significantly reduced NaASO 2 -induced promoter activity (Fig. 4B). This suggests that the putative reverse ELK-1-binding element located in the region Ϫ235 to Ϫ150 is necessary for transcriptional activation of the p21 promoter in response to NaASO 2 .
ELK-1 Directly Binds to the p21 Promoter-To determine whether ELK-1 binds to the p21 promoter, an EMSA was performed. Nuclear extracts from HaCaT cells were incubated with the radiolabeled oligonucleotides whose sequences corresponded to the ELK-1-binding sequence found between nucleotides Ϫ190 and Ϫ170 of the p21 promoter. As shown in Fig.  4C, oligonucleotides containing this ELK-1-binding motif formed protein-DNA complexes, which were competed out by the addition of unlabeled oligonucleotide probe. The specificity of ELK-1 binding was confirmed by the failure of a radiolabeled probe carrying a mutation in the ELK-1-binding site core sequence (TTCC 3 TTGG) to form protein-DNA complexes. To verify the binding of ELK-1 to the p21 promoter at the chromatin level, we cross-linked DNA and bound proteins in NaASO 2 -treated HaCaT cells using formaldehyde. Cross- . C, nuclear extracts from HaCaT cells treated with 50 M NaASO 2 for 15 min were probed with 32 P-labeled oligonucleotides with sequences corresponding to the region of the p21 promoter containing the ELK-1-binding site (Ϫ235 to Ϫ151 bp) in wild type or mutant (mtElk-1). To compete out labeled probes, unlabeled wild-type oligonucleotide (Competitor) was added in 10-and 100-fold excess. Arrow, DNA-ELK-1 complexes; arrowheads, nonspecific bindings. D, HeLa cells treated with NaASO 2 for 15 min were cross-linked, lysed, and immunoprecipitated with anti-phospho-ELK-1 (Ser-383) antibody or normal rabbit IgG (negative control). Precipitated DNA was analyzed by standard PCR using primers specific for the target region (Ϫ255 to Ϫ114) or off-target region (Ϫ2235 to Ϫ2026). One aliquot of input DNA was used as a positive control.

ELK-1 Regulation of p21 and BAX Transcription
linked DNA-protein complexes were subjected to chromatin immunoprecipitation using a rabbit anti-phospho-ELK-1 antibody or normal rabbit IgG. The resulting immunoprecipitated DNA was amplified by PCR using primers designed to the promoter region (Ϫ255 to Ϫ89) of the p21 gene. Input genomic DNA was used as a positive control. As shown in Fig. 4D, a noticeable increase in the amount of protein-bound DNA in NaASO 2 -treated cells was detected using the anti-phospho-ELK-1 antibody but not normal rabbit IgG. The off-target region (Ϫ2235 to Ϫ2026) was not amplified, although positive results were obtained from input genomic DNA. These data indicate that ELK-1 physically interacts with the p21 promoter in vivo.
ERK Mediates NaASO 2 -induced Activation of ELK-1-ELK-1 is phosphoactivated following the activation of multiple MAPK pathways in response to various extracellular stimuli (39 -43). To determine the involvement of MAPK pathways in NaASO 2 -induced p21 expression, serum-starved HaCaT cells were treated with NaASO 2 for various periods of time, and the activation of three major MAPKs was measured using phosphospecific antibodies. Levels of phosphorylated ERK1/2, JNK1/2, and p38 MAPK increased rapidly but transiently in response to NaASO 2 treatment, whereas the overall levels of these proteins remained unchanged (Fig. 6A), suggesting the activation of these MAPK pathways by NaASO 2 . To identify the MAPK pathway responsible for NaASO 2 -induced activation of ELK-1 in HaCaT cells, the effects of chemical inhibitors of MAPK signaling on NaASO 2 -induced ELK-1 phosphorylation were studied. All three MAPK inhibitors tested (the MEK1 inhibitor U0126, the p38 inhibitor SB203280, and the JNK inhibitor SP600125) strongly inhibited the ability of NaASO 2 to induce phosphorylation of ELK-1 on Ser-383 (Fig. 6B). To further determine the contribution of NaASO 2 -stimulated MAPK signaling to ELK-1 trans-activation, HaCaT cells were transfected with Gal4-Elk1/pFR-Luc trans-acting reporter constructs, along with constructs encoding mutant forms of MAPK signaling molecules. In line with the results obtained using chemical inhibitors, transient expression of DN-MEK1, DN-
To determine whether these MAPKs are functionally linked to NaASO 2 -induced p21 expression, we used Western blotting to examine the effects of chemical inhibitors on the accumulation of p21 protein. Interestingly, pretreatment with the MEK inhibitor U0126, but not the JNK inhibitor SP600125 or the p38 kinase inhibitor SB203580, abrogated the ability of NaASO 2 to induce the accumulation of p21 protein (Fig. 6D). Furthermore, transient expression of either DN-MEK1 or DN-ERK2 efficiently attenuated NaASO 2 -induced activation of the Ϫ235/ϩ38 construct of the p21 promoter (Fig. 6E). Collectively, although all three MAPK pathways can activate ELK-1, it seems that only the ERK pathway is critical for NaASO 2 -induced activation of p21 transcription.
Expression of ELK-1 siRNA Attenuates NaASO 2 -induced Expression of p21 and Apoptosis-We used RNA interference to test whether silencing endogenous ELK-1 expression reduces p21 expression. When HaCaT cells were transiently transfected with ELK-1 siRNA, along with the Ϫ235/ϩ38 construct of the p21 promoter, NaASO 2 -induced reporter activity was significantly attenuated (Fig. 7A). To further probe the involvement of ELK-1 in NaASO 2 -induced p21 expression, we established cell lines stably expressing ELK-1 siRNA (HaCaT/ siElk-1) or scrambled siRNA (HaCaT/Cont). Stable knockdown of ELK-1 by siRNA was evaluated by Western blotting (Fig. 7B). Silencing endogenous ELK-1 substantially attenuated the abil-ity of NaASO 2 to induce p21 expression. Interestingly, we found that NaASO 2 -induced BAX expression was also reduced in HaCaT/siElk-1 cells. In addition, HaCaT/siElk-1 cells displayed resistance to NaASO 2 -induced apoptosis and a decline in the G 1 population (as compared with control cells ; Fig. 7C). These data identify ELK-1 as the transcription factor responsible for NaASO 2 -induced up-regulation of p21 gene expression and BAX expression.
EGR-1 Functions Downstream of ELK-1 to Activate BAX Expression-EGR-1 can directly trans-activate the BAX promoter (56). EGR-1 is a known ELK-1 target and strongly induced by NaASO 2 . Because no ELK-1-binding elements have been identified in the BAX gene promoter region, we hypothesized that suppression of BAX expression in HaCaT/siElk-1 cells might be mediated by EGR-1. To test this possibility, we investigated the possible involvement of EGR-1 in NaASO 2induced BAX expression. As expected, the ability of NaASO 2 to induce EGR-1 expression was substantially attenuated in HaCaT/siElk-1 cells compared with HaCaT/Cont cells (Fig.  8A), indicating that EGR-1 expression is regulated by ELK-1. Forced expression of EGR-1 in HaCaT cells activated the BAX promoter in a plasmid concentration-dependent manner (Fig.  8B). Next, we examined whether the EGR-1-binding sequence in the BAX gene promoter is necessary for NaASO 2 -induced trans-activation. We showed that site-directed mutation of the EGR-1-binding core sequence within the BAX promoter (acaagcctGGGcgtggg 3 acaagcctTTTcgtggg) significantly attenuated luciferase reporter activation by NaASO 2 (Fig. 8C). These data suggest that activation of the BAX promoter by

NaASO 2 in HaCaT cells involves ELK-1-mediated EGR-1 expression.
To confirm the functional role of EGR-1 in BAX expression, we generated HaCaT/siEgr-1 cells, which stably express EGR-1 siRNA, and determined the effect of stable knockdown of endogenous EGR-1 protein on BAX expression. As shown in Fig. 8D, stable knockdown of EGR-1 expression prevented the ability of NaASO 2 to induce BAX, whereas p21 expression was not affected. To further probe the involvement of EGR-1 in NaASO 2 -induced BAX expression, we prepared primary mouse embryonic fibroblasts (MEFs) from Egr-1 wild-type (ϩ/ϩ) and Egr-1 knock-out (Ϫ/Ϫ) mice. The induction of BAX expression by NaASO 2 was greatly reduced in Egr-1 Ϫ/Ϫ MEFs, whereas p21 expression was not affected as compared with Egr-1 ϩ/ϩ cells (Fig.  8E). Given that MEFs contain wild-type p53, it is likely that EGR-1 mediates NaASO 2 -induced BAX expression in a variety of cell types, regardless of their p53 expression.

DISCUSSION
Epidemiological studies have shown that long term exposure to low concentrations of arsenite is associated with an increased risk of human cancers, including those of the skin, respiratory tract, hematopoietic system, and urinary bladder (57). Based on this information, the International Agency for Research on Cancer and the United States Environmental Protection Agency classified arsenite as a human carcinogen. The general population is exposed to arsenic through the air, soil, drinking water, food, and beverages. The amount of ingested arsenic appears to be dependent upon living environment, life style, and dietary patterns. However, the relationship between the dose of ingested arsenite and the cumulative concentrations in the body is currently unknown. The effect of low arsenite concentrations on the transformation of cells has been well studied (14 -16); however, the cytotoxic mechanism induced by high arsenite doses is unclear.
In this work, we investigated the effect of high NaAsO 2 concentration on cytotoxicity using a p53-mutated HaCaT model cell system. Herein, we provide evidence that, in response to NaASO 2 treatment, ERK activates ELK-1, an ETS family transcription factor, which in turn trans-activates a putative cisacting response element within the p21 promoter to induce p21 expression (p53-and EGR-1-independently). Furthermore, we show that an ERK/ELK-1 cascade indirectly activates the BAX  JULY 29, 2011 • VOLUME 286 • NUMBER 30

ELK-1 Regulation of p21 and BAX Transcription
promoter via induction of EGR-1. We suggest that NaASO 2induced p21 and BAX expression is highly dependent on ERK/ ELK-1 signaling in p53-mutated HaCaT keratinocytes.
Our data show that exposing HaCaT cells to high concentrations of NaASO 2 (Ն50 M) inhibits the cell cycle and induces apoptosis. These responses may stem from up-regulation of p21 and BAX expression as well as down-regulation of cyclins D1, A1, and B. Because the induction of p21 by NaASO 2 in HaCaT cells preceded the down-regulation of other cell cycleregulatory proteins, we suggest that up-regulation of p21 and BAX expression may represent an important mechanism by which NaASO 2 causes cytotoxicity. Thus, this study focused on the mechanisms behind p53-independent p21 and BAX gene expression in NaASO 2 -exposed HaCaT cells.
UVB increases the expression of p53, as well as p21 and BAX, leading to apoptosis even in cells carrying mutations in both p53 alleles (45). We thus tested whether NaASO 2 activates p53 in HaCaT cells. We showed that neither the expression nor transcriptional activity of p53 was affected by NaASO 2 treatment. In addition, a p21 promoter construct lacking a p53binding site was not activated by NaASO 2 . We therefore concluded that p53 was inessential to NaASO 2 -induced p21 expression, at least in HaCaT cells. Because NaASO 2 up-regulates EGR-1 expression in HaCaT cells (54), and EGR-1 stimulates transcription of the p21 gene by binding to specific sequences within its promoter (48 -49, 55), we examined the possible involvement of EGR-1 in NaASO 2 -induced up-regulation of p21 gene expression. Unexpectedly, we found no evidence for the involvement of EGR-1 in the regulation of NaASO 2 -induced p21 expression in HaCaT cells; the Ϫ150/ ϩ38 p21 promoter construct, which contains the EGR-1 site, did not respond to NaASO 2 , whereas transient transfection of EGR-1 siRNA had no effect on NaASO 2 -induced p21 promoter activity.
To identify the cis-acting response element that mediates NaASO 2 -induced p21 gene expression, we performed 5Ј-deletion analysis of the p21 promoter. We found that the promoter region spanning positions Ϫ235 to Ϫ150 is indispensable to the regulation of NaASO 2 -stimulated p21 promoter activity in HaCaT cells. Inspection of this region revealed the presence of a putative ELK-1-binding core motif, 5Ј-TTCC-3Ј, complementary to the core motif, 5Ј-GGAA-3Ј, in the antisense strand. Through mutational analysis of the p21 promoter, we demonstrated that disruption of this core ELK-1 binding motif (TTCC 3 TTGG) completely abrogated NaASO 2 -induced activation of the p21 promoter. Furthermore, we showed that forced expression of ELK-1 itself enhanced p21 promoter activity and that the introduction of ELK-1-specific siRNA into HaCaT cells efficiently attenuated NaASO 2 -induced p21 promoter activity. Direct binding of ELK-1 to the p21 promoter was confirmed by EMSA and ChIP assay. These results strongly suggest that ELK-1 participates directly in NaASO 2 -induced activation of the p21 promoter. Our data also show that NaASO 2 activated three major MAPKs (ERK, JNK, and p38 kinase) in HaCaT cells. However, only the ERK pathway was critical to NaASO 2 -induced p21 expression in HaCaT cells, as revealed using chemical inhibitors and dominant negative MAPK mutant constructs. NaASO 2 induces p21 expression via p38 MAPK in NIH3T3 cells (20) and via JNK in human umbilical vein endothelial cells (21). Because ELK-1 is a well known target of three major MAPKs (39 -41), it may separately contribute to ERK-, JNK-, or p38-induced p21 expression, depending on the cellular context. . C, HaCaT cells were co-transfected with 0.2 g of BAX promoter construct (pBax-Luc(Ϫ478/ϩ4), -Luc(Ϫ297/ϩ4), or -Luc(Ϫ478/ ϩ4)mtEgr1). The core EGR-1-binding motif is enclosed in a box. After 48 h, cells were treated with 50 M NaASO 2 for 8 h, and luciferase activities were measured. Values for firefly luciferase were normalized to those for Renilla luciferase. Data represent the mean Ϯ S.D. of three independent experiments, each performed in triplicate (*, p Ͻ 0.01 versus untreated cells). D-E, HaCaT cells expressing scrambled control (Cont) or EGR-1 siRNA (siEgr-1) (D) and Egr-1 ϩ/ϩ or Egr-1 Ϫ/Ϫ MEFs (E) were treated with 50 M NaASO 2 for 1 or 24 h. Then total cell lysates were prepared and tested for the expression of EGR-1, BAX, and p21 by Western blotting. GAPDH was used as an internal control.
The ternary complex factor subfamily of ETS transcription factors, whose members include ELK-1, SAP-1, and SAP-2/ ERP/Net, has been implicated in the regulation of gene expression, including that of immediate early response genes, such as FOS and EGR-1, in response to a variety of extracellular signals, through cooperative interactions with serum response element-bound SRF (39 -41, 43). However, ELK-1 can trans-activate its binding elements in the absence of SRF, for example within the mouse cytosolic chaperonin subunit (Cctq) gene promoter (58). Indeed, a whole group of ELK-1 target genes are largely regulated in an SRF-independent manner (59). Because no serum response element site has been identified in the p21 promoter, the binding of ELK-1 to the p21 promoter provides a further example of ELK-1 controlling target gene expression without associating with SRF.
The transcription factor SP1 plays a role in ERK-mediated p21 transcription in various cell types, including nerve growth factor-treated PC12 cells (60), Ras-transformed NIH3T3 cells (61), alkylphospholipid-treated HaCaT cells (33), and arsenic trioxide-exposed A431 epidermoid carcinoma cells (62). ETS and C/EBP␤ (63) are also involved in ERK-dependent, p53independent expression of p21 expression in primary hepatocytes. Therefore, it is possible that the induction of p21 expression by ERK involves multiple cis-acting elements. However, it should be noted that NaASO 2 also trans-activated a Ϫ235/ϩ38 construct lacking two ETS-binding sites at Ϫ1574 and Ϫ1347 (64) and a C/EBP␤-binding site at Ϫ1924 (63), but not a Ϫ150/ ϩ38 construct containing multiple SP1 sites (Ϫ119/Ϫ77) and an AP2 site at Ϫ102 (65). Furthermore, the Ϫ235/ϩ38mtElk1 construct, which carries a mutated ELK-1-binding sequence, but intact SP1 sites, was barely activated by NaASO 2 . Thus, we suggest that the ETS, C/EBP␤, and SP1 sites, which can be activated by ERK signaling, might not be essential for NaASO 2 activation of the p21 promoter in HaCaT cells. Nonetheless, we do not preclude the possibility that these transcription factors do contribute to full activation of the p21 promoter by NaASO 2 . Because the tumor suppressor BRCA1 activates the p21 promoter in a p53-independent fashion via the proximal region between Ϫ143 and Ϫ93 (66 -67), and the interaction of ELK-1 with BRCA1 enhances growth suppression in breast cancer cells (68), ELK-1 may interact with multiple nuclear proteins to enhance transcriptional activity in the proximal region of the p21 promoter.
Silencing of ELK-1 expression by RNAi in HaCaT cells resulted in reduced p21 and BAX expression in response to NaASO 2 exposure and conferred resistance to NaASO 2 -induced apoptosis. Given that (i) no consensus ELK-1-binding motif has been identified in the BAX promoter (69), EGR-1 can directly trans-activate the BAX promoter (56), and (iii) both EGR-1 and BAX levels were reduced by ELK-1 silencing, it is likely that EGR-1 plays a role in the induction of BAX by NaASO 2 . To test this idea, we transiently transfected HaCaT cells with EGR-1 siRNA. We found that NaASO 2 -induced BAX promoter activity was dose-dependently abrogated by transfection with EGR-1 siRNA. Our observation that NaASO 2 -induced BAX expression was largely abolished in MEFs from Egr-1 knock-out mice and in HaCaT cells expressing EGR-1 siRNA (HaCaT/siEgr-1) further supports a role for EGR-1 in NaASO 2 -induced BAX expression. Because MEFs express wild-type p53 and HaCaT cells contain the p53 mutation, it seems likely that EGR-1 activates BAX expression irrespective of p53 status. Because EGR-1 is up-regulated by the ERK/ELK-1 pathway, it appears that ELK-1 indirectly regulates BAX expression via EGR-1 in p53-mutated HaCaT cells.
Previous studies have shown that EGR-1 mediates radiationinduced apoptosis. For example, direct trans-activation of the BAX promoter in irradiated prostate cancer cells has been reported (56), suggesting that EGR-1 is proapoptotic under certain cellular conditions. However, although BAX expression was significantly reduced in HaCaT/siEgr-1 cells compared with HaCaT/Cont cells, the cleavage of caspase-2 and -7 (supplemental Fig. 1A) and apoptosis (supplemental Fig. 1B) were similarly induced by NaASO 2 treatment in both cell types. These findings suggest that EGR-1-mediated BAX induction is insufficient to induce NaASO 2 -induced apoptosis in HaCaT cells. Although the mechanisms of p21 apoptotic regulation are poorly understood, p21 can promote apoptosis under certain circumstances (70,71). For example, p21 overexpression enhances cisplatin-induced cell death in ovarian carcinoma cells (72), and the silencing of p21 by RNAi in HaCaT cells indicates that p21 functions in UVA-induced apoptosis and the G 1 /S phase cell cycle arrest (73). Thus, we suggest that the accumulation of p21 may play an important role in NaASO 2 -induced cytotoxicity through cell cycle dysregulation and apoptosis in HaCaT cells. Further study is required to determine the mechanism of p21-induced apoptosis.
In summary, our present study reveals that NaASO 2 -induced up-regulation of p21 and BAX expression was mediated by an ERK/ELK-1 cascade in p53-mutated HaCaT cells. ELK-1 directly transactivates the p21 gene promoter via a specific cisacting element and indirectly stimulates the BAX gene via induction of EGR-1. We conclude that NaASO 2 -induced cytotoxicity in p53-mutated HacaT cells is highly dependent on ERK/ELK-1 signaling, further extending our understanding of the regulatory mechanism by which MAPK signaling contributes to cellular cytotoxicity.